Munich Personal RePEc Archive
The dynamic effects of infectious disease outbreaks: the case of pandemic
influenza and human coronavirus
Verikios, George
KPMG Australia, Griffith University
June 2020
Online at https://mpra.ub.uni-muenchen.de/104434/
MPRA Paper No. 104434, posted 13 Dec 2020 20:45 UTC
George Verikios KPMG Economics
and
Department of Accounting, Finance and Economics Griffith University
Abstract
Pandemic influenza is a regularly recurring form of infectious disease; this work analyses its economic effects. Like many other infectious diseases influenza pandemics are usually of short, sharp duration. Human coronavirus is a less regularly recurring infectious disease. The human coronavirus pandemic of 2019 (COVID119) has presented with seemingly high transmissibility and led to extraordinary socioeconomic disruption due to severe preventative measures by governments. To understand and compare these events, epidemiological and economic models are linked to capture the transmission of a pandemic from regional populations to regional economies and then across regional economies. In contrast to past pandemics, COVID119 is likely to be of longer duration and more severe in its economic effects given the greater uncertainty surrounding its nature. The analysis indicates how economies are likely to be affected due to the risk1modifying behaviour in the form of preventative measures taken in response to the latest novel pandemic virus.
JEL codes: C68, E32, F15, I18.
Keywords: computable general equilibrium, human coronavirus, COVID119, infectious diseases, pandemic influenza, periodicity, trade linkages.
! "#
Thanks are due to Kevin Hanslow and two referees for helpful comments on this paper. Thanks are also due to Marc Mariano for research assistance. The views expressed here are the author’s and do not necessarily reflect those of KPMG or Griffith University.
$% "& ' "(
Infectious diseases are a leading cause of death worldwide accounting for a quarter to a third of all mortality.1 Despite developments in pharmaceuticals infectious disease rates are rising due to changes in human behaviour, larger and denser cities, increased trade and travel, the inappropriate use of antibiotic drugs, and the emergence of new and resurgent pathogens (Dobson and Carper, 1996). The current outbreak of human coronavirus (COVID119) is a reminder of the ease with which infectious disease outbreaks can cross borders and threaten economic stability. This has been observed before with similar outbreaks such as HIV, H1N1, H5N1 and SARS (e.g., Chou et al., 2004; Fan, 2003; Hai, et al. 2004; Keogh1Brown et al., 2010; Verikios et al., 2012). The threat to stability derives from a number of features of these outbreaks.
One, by definition emerging diseases are not commonly encountered by physicians and are thus capable of generating widespread infection and mortality prior to identification of the etiologic agent (e.g., HIV/AIDS). COVID119 has been especially challenging to contain because it is more contagious than influenza due to its ability to survive on surfaces and objects and transmit between people in this way (Centers for Disease Control and Prevention, 2020). Two, drug development and approval timeline lag well behind the emergence of these diseases such that the initial infection can result in significant mortality. Three, the constant adaptation of microbes, along with their ability to evolve and become resistant to antibacterial and antiviral agents, ensures that infectious diseases will remain a continuous and evolving threat.
Pandemic influenza is a regularly recurring form of infectious disease. It is possible that the current outbreak of human coronavirus may also become a regularly recurring form of infectious disease given it represents the second strain of severe acute respiratory syndrome coronavirus (SARS1CoV12) — the first strain led to the outbreak of SARS (SARS1CoV11) during 200212003 (Coronaviridae Study Group of the International Committee on Taxonomy of Viruses, 2020). This work assesses the economic effects of such infectious diseases with a focus on the dynamic effects. This is done by initially analysing a hypothetical influenza pandemic with characteristics consistent with previous historical occurrences; subsequently a pandemic that mimics some of the known (at this point) characteristics of the COVID119 pandemic is analysed.
There are a number of previous studies analysing the economic effects of global pandemics. A general equilibrium approach is the ideal framework for properly evaluating
1 See World Health Organization (2004, p. 121, Annex Table 2).
the economic impacts of public health emergencies such as pandemic influenza and human coronavirus. A partial equilibrium approach that only focuses on the health sector and forgone incomes resulting from disease1related morbidity and mortality, while ignoring effects in other parts of the economy (e.g., Sander et al., 2009), is incomplete. Illness and death due to public health emergencies raise perceptions of risk, leading to risk1modifying behaviour (such as prophylactic absenteeism from work and public gatherings) in an effort to reduce the risk of contracting illness. Risk1modifying behaviour affects consumption and reduces labour productivity. Deaths due to illness reduce the supply of workers. The effects of risk1modifying behaviour and deaths will affect all parts of the economy to a greater or lesser extent.
An important feature of pandemics is their short, sharp nature; they usually begin and end within a year. Previous studies focusing on the economic effects of global pandemics apply either a macroeconomic or computable general equilibrium (CGE) approach. Although macroeconomic (i.e., single sector) approaches (e.g., Keogh1Brown et al., 2010) have the advantage of applying quarterly models that allow them to capture the short, sharp nature of pandemics; they have the disadvantage of a single1sector approach that ignores sectors that are particularly relevant to the study of the economic effects of epidemics (e.g., medical services, international tourism). Although CGE (i.e., multi1sector) approaches (i.e., Chou et al., 2004) apply models that have the advantage of identifying multiple sectors in the economy, they have the disadvantage of annual periodicity and so are unable to capture accurately the short, sharp nature of influenza pandemics. The disadvantage of the CGE approach has been addressed by studies applying CGE models with quarterly periodicity, e.g., Verikios et al. (2012). The same approach is taken here.
The results show that pandemics can cause large short1term effects on output with small ongoing longer1term effects. The two main drivers of the contractions in output are the fall in international and domestic travel and tourism and reduced labour productivity, the latter being more important when severe mitigation measures are adopted by governments such as during the current coronavirus pandemic.
)% * +"'& , ( , ' -+ .+ !( # )%$ / &/(
Influenza is a contagious disease that causes seasonal epidemics globally and is a leading cause of infectious disease1related deaths in most countries. Influenza mortality is highly variable from year to year and is a contributor to the variability in the annual mortality rate of industrialised countries. In non1pandemic years, influenza typically kills hundreds of
thousands of people worldwide. Occasionally, influenza rates can reach pandemic proportions. There have been four influenza pandemics since the beginning of the 20th century — 1918, 1957, 1968 and 2009 — each of which was the result of a major genetic change to the virus.
Of these four influenza pandemics the most severe was the 1918 Spanish Flu pandemic where influenza mortality reached as high as 35 times the yearly average. More recently, May 2009 saw the emergence from Mexico of a new H1N1 virus capable of human1to1human transmission. The 2009 H1N1 virus was a novel type of influenza A that was highly transmissible yet ultimately mild. It quickly spread around the world infecting 74 different countries in all six continents within five weeks. The rate of spread was far more rapid than previously observed and was enabled by high volumes of international air traffic. The World Health Organization declared a pandemic on June 11, 2009. Thus, the virus took less than two months to emerge from Mexico and travel to all parts of the globe. It ultimately reached more than 200 countries and infected hundreds of millions of people. The World Health Organization declared the pandemic over by the end of the year reflecting the brief lifespan of such viruses.
The current human coronavirus pandemic (COVID119) began in December 2019 in China and was caused by a novel virus (SARS1CoV12) that is the second strain of severe acute respiratory syndrome coronavirus; the first strain led to the outbreak of SARS (SARS1 CoV11) in East Asia during 200212003 (Coronaviridae Study Group of the International Committee on Taxonomy of Viruses, 2020). The current coronavirus outbreak was declared a pandemic in March 2020 (United Nations, 2020) and is currently in play. As of April 29, 2020 there are 3.1 million confirmed cases in 185 countries, 217,000 deaths and 935,000 recovered cases (Center for Systems Science and Engineering, 2020). The virus is thought to spread primarily through person1to1person contact via droplets produced when an infected person coughs, sneezes or talks. Infection may also occur via contact with contaminated surfaces and objects; this is a novel transmission compared to influenza1type viruses (Centers for Disease Control and Prevention, 2020).
One important metric of an infectious disease outbreak is the case fatality rate (CFR):
the ratio of total deaths to total cases. The COVID119 outbreak is still in play and until it ends it is difficult to provide a proper comparison to previous global outbreaks. For instance, Oke and Heneghan (2020) estimate a CFR prediction interval for COVID119 of 0.82%19.64% as at the end of April 2020. They note “Evaluating CFR during a pandemic is, however, a very hazardous exercise, and high1end estimates should be treated with caution as the H1N1 pandemic highlights that original estimates were out by a factor greater than 10.” This is
explained by Rajgor et al. (2020) “A major challenge with accurate calculation of the CFR is the denominator” (p. 1), which is affected by asymptomatic cases, cases with mild symptoms, or misdiagnosed cases; thus under1 and over1estimating the CFR. Their assessment is that despite its high transmissibility the coronavirus CFR appears lower than SARS (9.5%) but higher than seasonal influenza (0.1%) and H1N1 (0.1%).2 This would be consistent with the current lower1bound prediction of 0.82% of Oke and Heneghan (2020).
In socio1economic terms the defining features of COVID119 are (i) the almost complete global spread of the virus within four months, and (ii) the extensive socioeconomic disruption due to preventative measures taken by governments. An example of unprecedented preventative measures is the almost total shutdown of domestic and international travel due to internal and external border closures (Salcedo et al., 2020).
)%) * ,, " *'!+ 0 *+/( '&
Direct economic effects of illness resulting from influenza include increased healthcare expenditures and workloads. Indirect effects include a permanently smaller labour supply due to deaths, and increased absenteeism from work by sick workers and by workers wishing to reduce the risk of contracting illness in the workplace, i.e., prophylactic absenteeism.
Prophylactic absenteeism is one example of voluntary risk1modifying behaviour in response to a pandemic. Other examples are reduced domestic and international travel, and reduced public gatherings at sporting and other events. Non1voluntary risk1modifying behaviour may be imposed on workers with children by school closures intended to mitigate the spread of the virus (Beutels et al., 2008). Thus, some workers will be forced to take leave to care for young children. Workers who take paid leave from work, whether forced or voluntary, reduce their firm’s labour productivity (i.e., output per worker) unless other workers can fully replace output lost due to absenteeism. This may be difficult during an influenza pandemic because the virus will be widespread and while many workers may not present to the health system they are likely to be less productive than would otherwise be the case.
Fan (2003) argues that a pandemic will reduce business investment due to increased uncertainty and risk, leading to excess capacity. Similarly, consumer confidence will decline due to uncertainty and fear, leading to reduced spending as people elect to be homebound to reduce the probability of infection—this is another example of risk1modifying behaviour.
2 The World Health Organisation does not provide an estimate of the CFR for H1N1. A range of estimates exist from different studies. Wong et al. (2013) review 50 studies that estimated the CFR for H1N1. The highest of these were around 0.1%, which is similar to seasonal influenza.
Reduced consumer confidence may particularly affect services involving face1to1face contact (e.g., tourism, transportation and retail spending). James and Sargent (2006) argue that evidence from past pandemics suggests that it is mainly discretionary spending (e.g., tourism and transportation) that is reduced.
Fan (2003) argues that an epidemic does not need to be of high morbidity and mortality in order to exert a large psychological impact on attitudes to risk. e.g., the 2003 SARS epidemic. James and Sargent (2006) examine evidence from the SARS epidemic and argue that the only economic impact was on air travel to affected locations and related impacts on accommodation. Keogh1Brown and Smith (2008) perform a retrospective analysis of the economic impact of the 2003 SARS epidemic and find that the economic effects were mainly but not exclusively centred on East Asian regions, and that the effects went beyond air travel and accommodation.
The response by individuals and governments to coronavirus supports the argument made above by Fan (2003). It is likely (but not certain) that the ultimate coronavirus CFR will be much lower than SARS but higher than seasonal influenza. Despite this, the socio1 economic disruption of coronavirus is disproportional to its morbidity and mortality. The mitigation measures taken to contain the virus are expected to lead to a significant global contraction in the order of 3% for 2020 (Gopinath, 2020).
1% * !( !
To model the economic consequences of infectious disease outbreaks a global CGE model is applied. A formal presentation of the model is available in Verikios (2017); a largely descriptive presentation is presented below.
The model represents the world economy as multiple regions engaging in trade and investment: markets are perfectly competitive, industry technologies are linearly homogeneous, and traded goods are imperfectly substitutable. Formally, the model is represented in homogeneous form by nonlinear equations specifying behavioural and definitional relationships as
(
,)
Fi 2, (1)
where Fi are i (=1,...,m) continuous and differentiable functions, is a m×1 vector of endogenous variables and is a m×1 vector of exogenous variables. Typically, describes changes in economic structure and policy (e.g., tax rates, labour productivity) and can be used to perturb the model to simulate changes in .3
3 The model is implemented in GEMPACK (Harrison and Pearson, 1996).
1%$ "&+" !. &+ "* &3
The model represents the world economy as economic activity occurring within and across regional economies. A regional economy may be either a single country (e.g., France) or a country group (e.g., the European Union). Each region produces a distinct variety of each commodity that is imperfectly substitutable with the varieties produced by other regions.
Each regional commodity is produced from inputs of domestically1produced and imported commodities and five primary factors: skilled and unskilled labour, capital, land and natural resources (see Figure 1).
( '& #
( '& $% "&' "'& , ( '#"&3 .& ' "( " * 3
Labour and capital are perfectly mobile across industries within a region whereas land and natural resources are fixed in industry usage. Consequently, wages for each labour type and the user price of capital are uniform across industries but the rental prices of land and natural resources can vary across industries.
Regional commodities can be consumed as inputs to final demand of which there are four categories: investment, government consumption, household consumption and exports.
Exports aside, final demands use composite commodities that are constant elasticity of substitution (CES) combinations of the domestic and the imported variety of each commodity, similar to composite commodity inputs for industries (Figure 1). Composite commodity inputs to investment and government consumption are determined by CES production and utility functions, while household consumption is determined by a constant1differences1 elasticity utility function. A Cobb1Douglas utility function with variable scale and share parameters determines the allocation of regional income across government consumption, household consumption and saving. The model is calibrated using the GTAP database (Narayanan and Walmsley, 2008) aggregated to 27 regions and 30 sectors (see Table 1).
Industry output
CES composite of inputs
Domestic commodity 1
Imported commodity 1
Domestic commodity N
Imported commodity N
Skilled labour
Unskilled
labour Capital Land Natural resources CES composite
commodity 1
CES composite commodity N
CES primary factor composite
+0 $% ( # + # " &# ( ! +"+0+#
Region Sector
1. Australia 1. Agriculture
2. Rest of Oceania 2. Coal
3. China 3. Oil
4. Japan 4. Gas
5. Korea 5. Other minerals
6. India 6. Processed food
7. Indonesia 7. Beverages and tobacco products
8. Singapore 8. Textiles, wearing apparel
9. Rest of Asia 9. Leather, wood products
10. Canada 10. Paper products, publishing
11. United States of America 11. Petroleum, coal products
12. Mexico 12. Chemicals, rubber, plastics
13. Argentina 13. Other mineral products
14. Brazil 14. Metals, metal products
15. Rest of South & Central America, Caribbean 15. Motor vehicles and parts
16. France 16. Other transport equipment
17. Germany 17. Electronic equipment
18. Italy 18. Other manufacturing
19. Great Britain 19. Utilities
20. Rest of European Union 20. Construction
21. Rest of Europe 21. Wholesale & retail trade
22. Russia 22. Air transport
23. Former Soviet Union 23. Other transport
24. Turkey 24. Communication
25. Rest of Middle East, North Africa 25. Other financial services
26. South Africa 26. Insurance
27. Rest of Sub1Saharan Africa 27. Other business services 28. Recreation, other services 29. Government services 30. Dwellings
As mentioned above, the model can be used to observe changes in endogenous variables due to changes in exogenous variables. Thus changes in tax rates and labour productivity can be applied on a region1specific or industry1specific basis and the model will project changes in industry output, prices, international trade, household consumption and GDP, among other variables. The dynamic aspects of the model are described below. This means that the time path of changes in endogenous variables can also be observed. The responses projected by the model reflect a perturbation of the initial steady state with a given capital1labour ratio to a new steady state once a new capital1labour ratio is reached.
1%) '+ + 4'+&" & 3 3 +!( #
Annual models are well suited to analysing events that last for about a year or more.
But for events that have short and sharp effects, such as infectious disease outbreaks, a quarterly model is more appropriate as an annual model will smooth short1term effects leading to potential underestimation of disruption. For example, if a pandemic caused an 80%
loss of inbound international tourism within a particular quarter, then the adjustment path of the tourism industry would be quite different from a situation in which international tourism declined smoothly by 20% for a year.
Annual CGE models are commonly solved in recursive fashion.4 Recursive or sequenced dynamic models usually divide time into discrete intervals and economic variables are assumed to change at the end of each interval. Such models take the form
i
( )
=G , (2)
where and are vectors of endogenous and exogenous variables in a period, and
(
1,...,)
G ii = m are m differentiable and continuous functions. Computations can then be carried out according to
i
( )
= ∇G , (3)
where refers to changes from one period to the next and ∇G ii
(
=1,...,m)
is the vector of first1order partial derivatives of Gi.Calibration of (3) requires an initial solution (i.e., a database consistent with the equations in (3)) mainly representing annual flows (e.g., household consumption, exports, etc.) and a choice of parameter values (e.g., price and income elasticities). To apply (3) as a quarterly model there are two options. One option is to modify the initial solution to (3) (i.e., the database) to represent quarterly rather than annual flows (i.e., divide annual flows by four).5 Another option is to leave the initial solution unchanged and modify (3) so that represents quarterly changes in exogenous variables, e.g., population growth. The simplest way to do this is to divide by four thus assuming constant quarterly changes through the year. This second option also requires the addition of equations that handle quarterly accumulation of stock variables; this can be done by again assuming constant quarterly changes through the year. With both options will represent quarterly rather than annual endogenous variables.
1%1 +.("+ + '!' +"( + ( / #"! "
In a dynamic framework, capital accumulation is handled by a stock1flow equation linking capital stocks across periods that allows for investment (i.e., new capital) and depreciation of existing capital at a geometric rate:
1
t t t t
r r r r
K+ =K +I −D ; (4)
where Krt is the quantity of capital available for use in region r in year t, Irt is the quantity of new capital created (i.e., investment) in region r during year t, and Drt is depreciation of existing capital in region r. With Irt and Drt representing annual values, Krt in (4) will grow
4 The exceptions are intertemporal models that compute results simultaneously for all time periods, e.g., McKibbin and Wilcoxen (1999).
5 See Kouparitsas (1998) for an example of this approach.
at an annual rate. The model contains a second stock1flow relationship that treats accumulation on a quarterly basis:
1
q q q q
r r r r
K + =K +I −D . (5)
In deriving (5), quarterly values for depreciation, Drq, and investment, Irq, are applied that ensure Krq accumulates at a quarterly rate. That is, Drq=Drt 4 and Irq=Irt 4. The relevant equation for any given simulation depends on whether annual or quarterly dynamics are applied.
The investment1capital ratio, Irq Krq, in each region is assumed to be a positive function of the rate of return. This relationship is calibrated to reflect increasing (convex) costs of adjusting the capital stock by larger amounts. This treatment captures the inertia in investment behaviour as observed in empirical studies.6
1%1 * +0 '& !+& "
The supply of each labour type is assumed to be sensitive to the real wage consistent with international evidence on non1zero wage elasticities of labour supply (e.g., Bargain et al., 2011):
( )
t
t t
lr
lr lr
t r
LS RW A
POP
= β . (6)
Thus the supply of labour type l in region r in year t, LSlrt , as a share of population in year t,
t
POPr, is determined as a function of the real post1tax wage received by labour type l in region r in year t, RWlrt. Alrt and β are positive constants. With the uncompensated labour supply elasticity β = 0.15, labour supply is only slightly responsive to the real return to labour in each region. Note that a similar relationship to (6) applies in the quarterly model.
Employment by labour type is an endogenous function of slowly1adjusting wage rates.
In the short1run, wage rates grow at the rate of inflation from period to period in the absence of labour market shocks. Where the market is in disequilibrium (i.e., the unemployment rate deviates from its long1run value) wage rates will grow more (less) quickly than the inflation rate if the unemployment rate is falling (rising). Hence, the rate of wage adjustment per period depends on labour market conditions relative to the long1run unemployment rate. This is consistent with empirical evidence that the rate of wage adjustment is partially determined by non1Walrasian features of the economy (see Romer (2001), chapter 9).
6 See Chirinko (1993) for a survey of approaches to modelling inertia in investment behaviour.
1%5 6 &+"( ! & #' "#
The baseline path of the model begins from an observed equilibrium (2004) that evolves to a balanced growth path. The starting data are updated from 2004 to 2009 by incorporating historical movements in real GDP and labour supply.7 From 2010 to 2025 historical population movements are applied, and 2% annual growth in labour productivity and fixed factors (land and natural resources) is assumed in all regions. Although the initial data do not represent a steady state, the assumptions for growth in labour productivity and fixed factors mean that the growth rate for all quantity variables converges close to the long1run growth rate for each region: population growth plus the growth in labour productivity.
7% * ( , "( '# (# +# !
The morbidity and mortality impact of the pandemic on the age cohorts of the population are taken from Verikios et al. (2011) who model them using a variation on the classic Susceptible1Exposed1Infected1Recovered (SEIR) model of infectious disease transmission. The SEIR model computes the theoretical number of people infected with an infectious disease in a closed population over time. This type of modelling is applicable to diseases where an individual that has recovered from the disease is removed from the susceptible population. In order to derive the equations of the model, the population of a single region or demographic group is divided into eight subpopulations: susceptible, vaccinated, exposed, three subsets of infected (untreated, hospitalised, and treated), recovered, and dead. Figure 2 displays the way the model replicates the dynamics of a pandemic.
The susceptible population is decreased through vaccination and exposure to the virus;
conversely, it is increased by the loss of vaccine1acquired immunity. Vaccinated individuals are not considered to be completely protected but become exposed at a rate much lower than the susceptible population. Vaccine efficacy can vary over the course of the epidemic.
Typically, around one month is required after vaccination until the individual has produced sufficient antibodies for the vaccine to be effective. Even after this initial time period, vaccine efficacy will be less than 100% due to varying individual antibody response to vaccine and the possibility of viral mutation or an imperfectly1matched vaccine.
7 The GDP data are sourced from the International Monetary Fund’s World Economic Outlook Database. The labour supply data are sourced from the International Labour Organization’s Labour Statistics Database.
( '& )% * !
After exposure, individuals progress to one of three infected states: untreated, treated, and hospitalised. The proportion of individuals progressing into each category is dependent on viral characteristics. As the virulence of the virus increases, the proportion of the infected receiving treatment also increases. The duration of infectiousness and transmission probabilities are decreased for those receiving treatment. Through this mechanism the parameters of the model allow for the simulation of behavioural and medical quarantine.
An increase in virulence can result in a reduction of average transmissibility;
individuals with a severe virus tend to be too ill to be out in the community transmitting the virus. In addition, those receiving treatment or who are hospitalised will have reduced contact and transmissibility due to precautionary measures such as masks, gloves, and isolation.
Individuals remain infectious during the entire course of their clinical infection. Once they have progressed out of the infectious state they can no longer transmit the virus to others.
After infection, individuals progress to one of two groups: recovered or dead. The rate at which individuals progress from one of the three infected states (hospitalised, treated, or untreated) to the end states is dependent on virulence and the level of treatment during infection. Hospitalised individuals have the highest death rate, followed by treated and untreated cases. Many of the untreated individuals are likely to have subclinical or asymptomatic infections, which reduces death rates despite the fact that some with untreated clinical infections may be more likely to die.
Those individuals progressing to the recovered state are considered to have immunity for the duration of the pandemic. Pandemics tend to come in waves and infect geographic areas for short periods of time. It is unlikely that in the case of an influenza pandemic the strain will mutate enough to cause re1infection during a single pandemic wave. There are five morbidity states included in the model:
1. subclinical, where the infected person seeks no medical attention but purchases pharmaceuticals;
2. physician and flu clinic, where the infected person seeks medical attention by visiting a physician or a flu clinic;
3. hospitalisation, where the infected person is hospitalised and survives;
4. intensive care unit (ICU), where the infected person is hospitalised and spends time in an ICU and survives; and
5. death, where the infected person is hospitalised and dies.
5% * .+ !( # +&( #
5%$ * ( , ' -+ .+ !(
The influenza pandemic applied here begins in Vietnam and has a global fatality impact of around 0.01%. It is an extremely transmissible influenza virus with a global attack rate of approximately 40%, despite the availability of an effective vaccine within months of the outbreak. The case fatality rate is 0.5%, which is similar to the 1957 influenza pandemic.
Unlike seasonal influenza where typically 90% of the fatalities are observed in individuals older than 65, this virus has an equal case fatality across ages consistent with the 2009 H1N1 pandemic. The majority of cases are subclinical or physician visits and approximately one1 fifth of those hospitalised require intensive care, analogous to what is observed in seasonal outbreaks.
The pandemic scenario has viral characteristics that are plausible but less extreme than historical events, such as the 1918 pandemic. Figure 3 shows how the pathogen characteristics of the scenario compares to historical flu pandemics. Epidemic curves were created for each country or region using the calculated correlation between the observed 2009 H1N1 pandemic transmission timeline and country density. A log1linear relationship between population and weeks until pandemic peak represented the most appropriate statistical fit.
The epidemic curves were used to develop weekly totals of individuals in each of the five morbidity classes by country or region.
5%$%$ &0( ("3 + ! &"+ ("3
The SEIR model provides estimates of the number of persons newly infected per week and the severity of their infection in each region over the course of the year. Overall infection rates are high relative to previous pandemics: see the x1axis in Figure 3. Further, infection rates are higher in lower income countries than in higher income countries, and higher in Asian regions, where the pandemic is assumed to begin, than in non1Asian regions.
( '& 1% +"* *+&+ " &(#"( # , "* ! .+ !( & +"(/ " *(#" &( + .+ !( #
Notes: Virulence is measured in terms of the case1fatality rate or deaths per infected case. Infectiousness (or transmissibility) represents the speed at which a pandemic will spread within a population and the total number of people that will be infected.
Infectiousness of influenza is measured by the basic reproductive rate (R0), or the mean number of secondary cases an infectious case will cause in a population without immunity and without intervention. R0 > 1 is necessary for an epidemic to occur.
The distribution of infected cases is skewed towards less severe cases and only a small proportion of infections result in death (0.5%). In historical terms, the pandemic’s virulence (i.e., initial deaths per case) is similar to previous pandemics as shown on the y1axis in Figure 3. Therefore, the majority of cases remain at the subclinical and physician/flu clinic levels of contact with the health system.
Figure 4 presents a picture of the dynamics of the influenza outbreak; it is typical of historical influenza pandemics. Globally, new infections occur over about two1thirds (about 35 weeks) of the year. New infections peak around week 16, and there are two further smaller peaks around weeks 22 and 32. The pattern of new global infections partially obscures the short, sharp nature of the outbreak at the regional level. Figure 4 shows that new infections last for only around 10 weeks in each region, on average. It also shows how the outbreak moves from continent to continent depending on international travel patterns. Thus, the outbreak begins in the Rest of Asia (Vietnam), moves quickly to China, then the United States and Africa, and lastly to South America. These dynamics are important in determining the timing of the economic effects across regions. The economic aspects of the outbreak’s dynamics can not be captured by an economic model that does not also reflect these dynamics, i.e., periodicity of much less than one year.
!
!
" !
!
!
" !
!
!
!
!
" !
!
!
" !
!
!
!
"#$###$###
%#$###$###
&#$###$###
'#$###$###
(#$###$###
)#$###$###
*#$###$###
" & ( * + "" "& "( "* "+ %" %& %( %* %+ &" && &( &* &+ '" '& '( '* '+ ("
, - , , . $ / , 0 ,
( '& 7% ( , "( # . & ( # " & ( + . .' +"( # 8 .+ !( ( , ' -+ # +&(
5%$%) (& " !( ,, "#
Previous analysis of influenza outbreaks and their potential economic effects highlights a number of channels through which an economy might be affected by a serious outbreak of influenza.8 These channels include: reduced consumption by households of tourism, transportation and retail trade; increased absence from the workplace due to illness or for prophylaxis; school closures; and higher demands for medical services. Considering these channels, four types of economic shocks are applied to simulate the influenza pandemic.
1. A temporary surge in demand for hospital and other medical services. Increased medical spending related to each pandemic is applied as increased expenditure on the Government services sector.
2. A temporary upsurge in sick leave and school closures. This represents workers falling ill and parents caring for children. These effects are modelled as a temporary decrease in output per worker (labour productivity).
3. Deaths with a related permanent reduction in the labour force.
4. Temporary reductions in international tourism and business travel. These are a function of the number of persons infected and the initial deaths per case. The shocks to tourism are applied to exports of four sectors: Wholesale and retail trade, Air transport, Other transport, and Recreation and other services. Such purchases represent spending by tourists.9
Note that with government consumption rising and household consumption falling, the national saving rate is not constant and may rise or fall depending on the relative changes in government and household consumption rates. Furthermore, the government budget is not held fixed in any period of the scenario.
8 See Chou et al. (2004), Fan (2003), James and Sargent (2006), Jonung and Roeger (2006), Keogh1Brown et al.
(2010), McKibbin and Sidorenko (2006) and Verikios et al. (2012).
9 The shocks are explained in detail in Verikios (2017).
5%) * *'!+ & +/(&'# .+ !(
Using the influenza scenario described above as a benchmark, a scenario is fashioned to represent what is known about the human coronavirus pandemic as at the end of April, 2020.
Using data from Roser et al., (2020) the spread of new infections is calculated from China in December, 2019 to other countries over the 17 week period ending April 29, 2020; these data are presented in figure 5. The data show that by week 9 four major countries and regions showed new weekly infections of greater than 1,000: China, Japan, Italy and Middle East1 North Africa. By week 11 another six more major countries and regions had 1,000 or more new weekly infections: the United States, France, Germany, the United Kingdom, Other European Union countries, and Other European countries. By week 14 all other major countries and regions were recording 1,000 or more new weekly infections: Oceania by week 12, Brazil by week 12, Other South and Central American countries by week 12, Russia by week 13, India by week 14, Mexico by week 14, and Indonesia by week 15. Argentina is the only G20 country that did not reach 1,000 or more new weekly infections by week 17.
# (#$###
"##$###
"(#$###
%##$###
%(#$###
" % & ' ( ) * 1 + "# "" "% "& "' "( ") "*
, 2 - 3 4
, - , .
, 0 , 5 6 7
4 5 8 7
. , ,
( '& 5% ( , "( # . & 03 & ( 8 & +/(&'#
For many of the countries and regions already discussed the rate of new weekly infections have already peaked and are now falling (as at end of April, 2020. Those still showing a rising rate of new weekly infections are India, Singapore, Other Asian countries, Canada, the United States, Mexico, Brazil, Other South and Central American countries, Russia, Other former Soviet Union countries, Middle East1North Africa, South Africa and Other sub1Saharan African countries.
In the absence of further data, assumptions are applied to project the path of new infections from week 18 onwards. This is done by assuming a 5% weekly geometric rate of decay. This gives a path of new infections until week 52 that varies by country and country
group. The direct effects of the coronavirus pandemic are then calculated as a function of the country1specific path of new infections through the year.
5%)%$ (& " !( ,, "#
Using the data discussed above, direct economic effects qualitatively similar to the influenza scenario are designed so that regions whose rate of new infections have peaked start to ease mitigation measures in quarter 3, 2020. In contrast, regions whose rate of new infections are yet to peak start to ease mitigation measures in quarter 4, 2020: this assumes that the rate of new infections in these regions peaks by July 2020. These assumptions are not intended as predictions or probable. They only serve to provide a temporal boundary around the epidemiological characteristics of the outbreak, which in turn allows the analysis of the possible economic impacts of the pandemic. Six types of economic shocks are applied to simulate the coronavirus pandemic.
1. A temporary surge in demand for hospital and other medical services. This effect is scaled to reflect the number of coronavirus infections relative to the number of pandemic influenza infections. Note that the number of coronavirus infections (as at the end of April 2020) are much lower than the number of pandemic influenza infections. This can be seen by comparing the scales in figures 4 and 5.
2. A temporary surge in demand for policing and related services. This effect is related to the enforcement of wide1ranging and compulsory social distancing measures imposed almost universally.10 It is assumed that this effect peaks at 5% of baseline government expenditures for all countries except Sweden and Singapore. The peak increase in spending coincides with the peak quarter of infections and then falls as new infections decline during the year.
3. A temporary upsurge in sick leave, school and university closures, workplace closures and cancelled public events and gatherings. These effects are designed to reflect the wide range of severe mitigation measures that countries have undertaken to enforce compulsory social distancing in efforts to contain the spread of coronavirus.11 These effects focus on the impact on workplaces. It is assumed that this effect peaks at 120% of baseline economywide labour productivity for all countries except for Sweden and Singapore. For Sweden and Singapore the peak is assumed to be 15%. The peak fall in labour productivity coincides with the peak quarter of infections and then falls as new infections decline during the year.
10 There seem to be only two known exceptions to compulsory social1distancing measures during the coronavirus pandemic: Sweden and Singapore (Townsend, 2020).
11 See New York Times (2020) for an extensive list of cancelled events.
4. Deaths with a related permanent reduction in the labour force. Similar to the point made in (1) above regarding infections, the number of coronavirus deaths (as at the end of April 2020) are much lower than the number of pandemic influenza infections.
5. Temporary reductions in domestic and international tourism and business travel. During the initial stages of the outbreak there was a fall in international travel and tourism related to uncertainty surrounding the epidemiological characteristics of the virus and the cancellation of international gatherings. Eventually most countries closed their borders to all visitors and only allowed the entry of returning citizens and residents.12 This has led to the almost total shutdown of international travel and tourism. This is imposed as global reductions in the relevant sectors of 25% to 60%. It is assumed that this peak effect is timed to coincide with the number of coronavirus infections through the year.
6. Temporary switching of household and business expenditure away from domestic tourism and travel, restaurants and accommodation, sporting events and in2store retail purchases.
The severe mitigation measures led to restrictions in a range of activities where person1to1 person contact is typical; this affected restaurants and accommodation, sporting events and in1 store retail purchases of non1essential goods. In most countries the mitigation measures also severely restricted inter1 and intra1city movement, thus severely affecting domestic tourism and travel. The switching of expenditure away from these commodities is imposed as a 50%
reduction in domestic sales in the relevant sectors. This peak effect is timed to coincide with the number of coronavirus infections through the year.
9% * !( ,, "# , .+ !( ( , ' -+
In this section we apply the direct effects of the pandemic to the economic model.
Figure 6 reports the effects on the levels of selected global variables relative to baseline;
Table 2 reports the real GDP effects for all regions.
12 See Salcedo et al. (2020) for a list of travel restrictions by country.
!(9#
!'9#
!&9#
!%9#
!"9#
#9#
"9#
%9#
%#"+:' %#%#:% %#%#:' %#%":% %#%":' %#%%:% %#%%:' %#%&:% %#%&:' %#%':% %#%':' %#%(:% %#%(:' 6 ;
<
( '& 9% 6 0+ ,, "# :. & "+ *+ ;
+0 )% <'+&" & 3 ! ,, "# 0+ + & ( + 6 :. & "+ *+ ;
Region Quarter
2020:1 2020:2 2020:3 2020:4 2021:1 2021:2 2021:3 2021:4
World 10.80 12.97 12.23 11.54 0.39 0.96 0.98 0.81
Australia 10.57 12.51 12.03 11.38 0.41 0.98 1.00 0.77
Rest of Oceania 10.54 12.27 11.85 11.35 0.26 0.82 0.82 0.73
China 11.51 15.68 14.58 13.59 0.36 1.95 2.18 1.79
Japan 10.84 12.59 11.90 11.21 0.53 0.99 0.96 0.75
Korea 11.13 13.62 12.83 12.00 0.27 1.16 1.39 1.26
India 10.09 10.74 10.56 10.31 0.31 0.42 0.41 0.35
Indonesia 10.70 12.50 11.94 11.41 0.25 0.93 1.10 1.02
Singapore 12.40 17.35 15.84 14.52 10.31 1.84 2.60 2.53
Rest of Asia 10.65 13.22 12.53 11.87 0.02 0.66 0.95 0.85
Canada 10.76 13.19 12.33 11.69 0.42 1.07 0.99 0.79
United States 10.52 11.86 11.24 10.64 0.48 0.52 0.40 0.32
Mexico 10.08 10.50 10.31 10.15 0.23 0.37 0.37 0.34
Argentina 10.57 12.24 11.75 11.25 0.12 0.63 0.75 0.62
Brazil 10.25 11.07 10.71 10.50 0.28 0.50 0.47 0.45
Rest of South America 10.42 11.99 11.49 11.04 0.30 0.69 0.71 0.56
France 11.01 13.83 12.78 11.91 0.60 1.30 1.25 0.96
Germany 11.06 13.92 12.85 11.96 0.50 1.26 1.29 1.04
Italy 10.54 12.21 11.73 11.26 0.23 0.70 0.77 0.64
Great Britain 11.21 14.40 13.28 12.29 0.63 1.52 1.49 1.16
Rest of EU 11.38 15.07 14.06 13.16 0.13 1.46 1.70 1.41
Rest Europe 10.84 13.07 12.26 11.50 0.39 0.96 1.07 0.92
Russia 10.46 12.05 11.71 11.31 0.05 0.65 0.89 0.90
Former Soviet Union 10.60 12.68 12.18 11.62 0.10 0.70 0.84 0.75
Turkey 10.10 10.74 10.59 10.38 0.04 0.16 0.27 0.30
Rest of Middle East 10.26 11.45 11.08 10.71 0.31 0.55 0.55 0.51
South Africa 10.51 12.13 11.56 11.12 0.26 0.71 0.70 0.60
Rest of Africa 10.41 12.37 11.87 11.48 0.13 0.67 0.65 0.52
The main impacts occur in 2020, the pandemic year. In quarter 1 of 2020 (2020:1) the pandemic causes global GDP to fall by 0.8% and labour by 1.2%. In 2020:2, GDP and labour fall further (13% and 14.2%) as new infections continue to rise (see Figure 4). As new infections peak in 2020:2, so too do the negative effects on GDP and labour. The movements in GDP are mostly driven by movements in labour. Labour is the only variable input in the short1run as capital stocks are subject to a one1period gestation lag and take time to respond to the pandemic. The contractionary effect of the pandemic reduces demand for labour by firms.
As wage rates are rigid in the short1run they do not fall enough to maintain labour usage, thus causing the fall in labour to be greater than would otherwise be case.
In 2020:3 the number of new global infections falls significantly as there are no new infections in Asia (the origin of the virus) and Europe; in 2020:4 there are only a small number of new infections and these are in Australia and South America. Although there are few new infections in 2020:3 and 2020:4, the pandemic keeps global economic activity depressed due to continuing precautionary measures (e.g., voluntary and compulsory restrictions on travel, prophylactic absenteeism by workers) to prevent the spread of the virus until no new infections are observed in a region. Nevertheless, economic activity begins to recover over 2020:314 as real wage rates respond with a lag to the fall in labour demand in 2020:1 and 2020:2. The delayed response of real wage rates begins to clear the excess supply in the labour market and so GDP and labour usage show smaller negative deviations from baseline in 2020:3 (12.2% and 12.9%) and 2020:4 (11.5% and 12%) compared to 2020:2.
To understand the effects of the four categories of shocks on global GDP, Figure 7 shows the individual effect of each shock. In 2020, the increase in medical services (due to hospitalisations and treatments), the reductions in labour supply (due to deaths) and labour productivity (due to lost workdays), and the fall in international tourism have negative effects on GDP. Of these shocks, the fall in international tourism dominates; the peak effect here is 12.8%, which accounts for most of the peak GDP effect of 13% in 2020 for all shocks combined. The importance of the tourism effects is also reflected in Figure 7 as large negative effects on global exports. The effect on exports is dominated by the fall in international travel and tourism due to the pandemic.
!&9#
!%9(
!%9#
!"9(
!"9#
!#9(
#9#
#9(
"9#
"9(
%#"+:' %#%#:% %#%#:' %#%":% %#%":' %#%%:% %#%%:' %#%&:% %#%&:' %#%':% %#%':' %#%(:% %#%(:'
( '& =% 6 0+ 6 8 ( (/( '+ #* # :. & "+ *+ ;
Figure 7 shows that the temporary reduction in international tourism dwarfs all other effects. This pandemic has a relatively high global infection rate in historical terms (24%).
As discussed earlier, the infection rate is the dominant determinant of the size of the tourism reductions. This characteristic of the pandemic causes the tourism effects of this pandemic to be very large relative to other effects. A pandemic with a much reduced infection rate (e.g., the 1957 pandemic) would be expected to lead to much smaller quarantine1like measures (e.g., social exclusion and travel restrictions) to prevent the spread of the virus, and thus much smaller temporary reductions in tourism.
Beginning in 2021:1, the exogenous shocks representing the pandemic are slowly withdrawn and this process is complete by 2021:4. Thus, Figure 6 shows GDP and labour move above baseline in 2021:1 by 0.4% and 0.7%. In 2021:2 and 2021:3, GDP and labour continue to move above baseline. This is due to the lagged response of wage rates to the end of the pandemic; this means that real wage rates are still 0.5% below baseline in 2021:4 as they only adjust slowly to the change in labour market conditions. From 2021:1 labour usage and GDP move back towards baseline as real wage rates also move back towards baseline.
By 2025 labour and real GDP return close to baseline levels.
Figure 8 presents the GDP effects for selected regions. All regions are projected to experience lower output in the short1run (Table 2): the differences across regions are purely due to the size of the negative output effects. Singapore experiences the largest negative deviation in 2020 real GDP: 17.3% in 2020:2. The very large effect on Singapore’s GDP is due to the strong negative effect on global trade (exports; see Figure 7) from the pandemic rather than the direct impact of reduced inbound tourism to Singapore. Singapore is an entrepȏt port through which large volumes of goods and persons pass on the way to their final destination. This is reflected in export and import to GDP shares of around 150% in the base data. This trade is heavily dependent on world trade but even more dependent on trade by Asian regions. As already noted, these regions are the most strongly affected by the influenza outbreak. Thus, Singapore is the region most exposed to the contraction in global and Asian trade due to its unique transit status.
Other countries with high trade1to1GDP shares also experience a strong negative indirect effect on their GDP due to the contraction in global trade. But the size of international tourism in GDP is also important for determining the GDP effects for other countries. Thus, China (15.7%), Great Britain (14.4%), Canada (13.2%) and Rest of Asia (1 3.2%) experience relatively large peak GDP effects (2020:2), whereas USA (11.9%), Brazil (1 1%) and India (10.7%) experience smaller peak GDP effects.
!19#
!*9#
!)9#
!(9#
!'9#
!&9#
!%9#
!"9#
#9#
"9#
%9#
&9#
%#"+:' %#%#:% %#%#:' %#%":% %#%":' %#%%:% %#%%:' %#%&:% %#%&:' %#%':% %#%':' %#%(:% %#%(:'
6 0
, -
- ,
( '& >% 6 ,, "# , & # " & ( # :. & "+ *+ ;
=% * !( ,, "# , .+ !( & +/(&'#
Here we apply the direct effects of coronavirus to the economic model. These direct effects are orders of magnitude larger than those applied earlier for the influenza pandemic due to the widespread and severe mitigation measures adopted in response to coronavirus. A notable effect of these responses is the extensive underutilisation of physical capital, i.e., capital idling. To account for this phenomenon the economic model is modified to allow for variable capital utilisation; the typical assumption is full capital utilisation. Allowing for capital idling means that large demand1driven reductions in output can be captured in the absence of large reductions in prices or deflation. This imposes consistency in the representation of the capital and labour markets (see section 3.3) in that neither market is assumed to clear on an ongoing basis.
In representing capital idling here the approach follows the treatment in real business cycle models where capital use is typically varied along the intensive margin (King and Rebelo, 1999). Here this is represented formally as
t t
r r t
t t r
r r
KU R
K P F
=
; 0 1
t r t r
KU K
< ≤
, (7)
where Krt is the quantity of capital available for use in region r in year t (see (4)) and KUrt is the quantity of capital in use. Thus,
t r t r
KU
K is the capital utilisation rate. This rate is a positive function of the ratio of the rental price of capital Rrt and average supply price for all industries
t
Pr , i.e., the real cost of capital. Frt is positive parameter.
Linearisation of (7) yields Crt =Rrt−Prt+Frt where Crt is the proportional change in the capital utilisation rate. Thus, a rising real cost of capital in the current period signals to firms to increase their use of currently idle capital rather than undertake more investment that would
